Miscellaneous fixes

src/access/esmf_trace/config.py

@@ -103,7 +103,7 @@ class PostSummarySettings:
 @dataclass(frozen=True)
 class PostRunSettings:
     name: str
-    output_index: list[str] | None = None
+    output_index: list[int] | None = None
     model_component: list[str] | None = None
     pets: list[int] | None = None
     stats_start_index: int | None = None
@@ -266,9 +266,9 @@ def load_yaml_config(config_path: Path, kind: config_kind):
                     output_index=output_index,
                     model_component=_norm_model_component(item.get("model_component", defaults.model_component)),
                     pets=pets,
-                    stats_start_index=_norm_int_or_none(item.get("stats_start_index", default.stats_start_index)),
-                    stats_end_index=_norm_int_or_none(item.get("stats_end_index", default.stats_end_index)),
-                    save_json_path=_norm_path_or_none(item.get("save_json_path", default.save_json_path)),
+                    stats_start_index=_norm_int_or_none(item.get("stats_start_index", defaults.stats_start_index)),
+                    stats_end_index=_norm_int_or_none(item.get("stats_end_index", defaults.stats_end_index)),
+                    save_json_path=_norm_path_or_none(item.get("save_json_path", defaults.save_json_path)),
                 )
             )
         return defaults, post_runs

src/access/esmf_trace/library.py

@@ -48,7 +48,21 @@ def post_summary_from_config(
         assert isinstance(defaults, PostSummarySettings)
     else:
         defaults = PostSummarySettings(**config_path["default_settings"])
-        runs = [PostRunSettings(**r) for r in config_path["runs"]]
+
+        merged_runs: list[PostRunSettings] = []
+        for r in config_path["runs"]:
+            rr = dict(r)
+
+            # inherit defaults if not explicitly set per-run
+            rr.setdefault("model_component", defaults.model_component)
+            rr.setdefault("pets", defaults.pets)
+            rr.setdefault("stats_start_index", defaults.stats_start_index)
+            rr.setdefault("stats_end_index", defaults.stats_end_index)
+            # do not inherit default combined save_json_path into per-run save_json_path
+            rr.setdefault("save_json_path", None)
+
+            merged_runs.append(PostRunSettings(**rr))
+        runs = merged_runs
 
     if post_overrides:
         defaults = replace(defaults, **dict(post_overrides))

src/access/esmf_trace/postprocess.py

@@ -42,7 +42,7 @@ def _slice_per_series_iloc(
     """
     Slice rows per (group) using iloc[start:end] in each group after sorting by order_cols
     If both start and end are None -> no slicing (full series).
-    If end is None -> no slicing (full series).
+    If end is None -> slice from start to end of series.
     """
     if start is None and end is None:
         return df
@@ -142,7 +142,7 @@ def _summarise_case(
         tmin=("duration_s", "min"),
         tmax=("duration_s", "max"),
         tavg=("duration_s", "mean"),
-        tmedian=("duration_s", lambda x: x.quantile(0.50)),
+        tmedian=("duration_s", "median"),
         tstd=("duration_s", "std"),
     ).reset_index()
 
@@ -174,21 +174,19 @@ def _summarise_case(
     )
     per_output = per_output.sort_values(["__case_name", "__output_index", "model_component"], kind="mergesort")
 
-    combined_by_comp = (
-        per_output.groupby(["__case_name", "model_component"], sort=False, dropna=False)
-        .agg(
-            hits=("hits", "mean"),
-            tmin=("tmin", "min"),
-            tmax=("tmax", "max"),
-            tavg=("tavg", "mean"),
-            tmedian=("tmedian", "mean"),
-            tstd=("tstd", "mean"),
-            pemin=("pemin", "min"),
-            pemax=("pemax", "max"),
-            ncpus=("ncpus", "mean"),
-        )
-        .reset_index()
-    )
+    grp_comp = ts.groupby(["__case_name", "model_component"], sort=False, dropna=False)
+    combined_by_comp = grp_comp.agg(
+        hits=("duration_s", "count"),
+        tmin=("duration_s", "min"),
+        tmax=("duration_s", "max"),
+        tavg=("duration_s", "mean"),
+        tmedian=("duration_s", "median"),
+        tstd=("duration_s", "std"),
+        ncpus=("pet", "nunique"),
+        pemin=("pet", "min"),
+        pemax=("pet", "max"),
+    ).reset_index()
+
     combined_by_comp["__output_name"] = "combine"
     combined_by_comp["__row_label"] = (
         combined_by_comp["__case_name"] + "/combine/" + combined_by_comp["model_component"].astype(str).str.strip()
@@ -197,7 +195,7 @@ def _summarise_case(
     return pd.concat([per_output[output_cols], combined_by_comp[output_cols]], ignore_index=True)
 
 
-def _resolve_save_json_path(save_json_path: str | None) -> Path | None:
+def _resolve_save_json_path(save_json_path: str | Path | None) -> Path | None:
     if save_json_path is None:
         return None
     p = Path(save_json_path).expanduser()
@@ -211,6 +209,8 @@ def post_summary_from_yaml(
     defaults: PostSummarySettings,
     runs: list[PostRunSettings],
     save_json_path: str | None = None,
+    include_combined: bool = True,
+    include_per_output: bool = True,
 ) -> pd.DataFrame:
     post_base_path: Path = Path(defaults.post_base_path)
     timeseries_suffix: str = defaults.timeseries_suffix
@@ -239,7 +239,7 @@ def post_summary_from_yaml(
             continue
 
         # Save per-run json if this run specified a save path (strict .json)
-        per_run_save = _resolve_save_json_path(r.save_json_path)
+        per_run_save = _resolve_save_json_path(r.save_json_path) if r.save_json_path is not None else None
         if per_run_save is not None:
             (
                 case_summary.reset_index(drop=True).to_json(  # ensure a clean row index
@@ -256,6 +256,12 @@ def post_summary_from_yaml(
     # Build combined table across all selected runs
     combined_df = pd.concat(per_case_tables, ignore_index=True)
 
+    if not include_combined:
+        combined_df = combined_df[combined_df["__output_name"] != "combine"]
+
+    if not include_per_output:
+        combined_df = combined_df[combined_df["__output_name"] == "combine"]
+
     wanted_cols = ["__row_label", "hits", "tmin", "tmax", "tavg", "tmedian", "tstd", "pemin", "pemax"]
     combined_df = combined_df.loc[:, [c for c in wanted_cols if c in combined_df.columns]]